In 2005 two well-known human geneticists, Francis Collins and Thomas Gelehrter, made a bet: Collins wagered that by the 2008 American Society for Human Genetics meeting, genomewide association studies would have led to the discovery of at least four “validated – not just guessed at” susceptibility variants for at least five common diseases.
Collins won his bet (and hopefully his beer) – by a margin of more than 200 variants. According to the Human Genome Epidemiology Network, there have been 328 genomewide association studies published to date, and there is no sign of a slow down (as any devotee of SNPwatch knows). This type of research has clearly been successful. But how useful are the variants they identify?
There are two directions from which to consider this question. One is to ask whether the common variants found in genomewide association studies add to science’s understanding of the biology of various common diseases, as well as medicine’s ability to treat them. The other is to ask whether the results of genomewide association studies have made it possible to predict individual genetic risk for various diseases. Three opinion pieces, each with a different perspective, address these questions in the New England Journal of Medicine this week.
David Goldstein, director of the Center for Human Genome Variation at Duke’s Institute for Genome Sciences and Policy, doesn’t doubt the veracity of variants found in genomewide association studies. But he does think that it’s time for geneticists to take a new approach.
“I believe attention should shift from genome scans of ever larger samples to studies of rarer variants of larger effect,” he writes in his NEJM commentary.
Goldstein says continued use of genomewide association approaches that look for common variants associated with common diseases and traits will ultimately lead to an unwieldy number of variants, each with vanishingly small effect sizes. For example, after making some assumptions he calculates that approximately 93,000 common variants could be needed to explain 80% of the population variation in height.
But variants with small effects can be biologically informative, argues Joel Hirschhorn, a Harvard genetics professor, in the second NEJM Perspective. For example, genomewide association studies have identified variants associated with type 2 diabetes, high cholesterol and low bone density that are located in genes targeted by drugs already approved by the FDA for the treatment of these conditions.
“Each of the associated variants at a drug-target locus explains less than 1% of phenotypic variation in the population, demonstrating that small effect sizes do not preclude biologic importance,” he writes.
The discovery of variants with small effect sizes can also point scientists in the direction of new avenues for drug research, says Hirschhorn. Genomewide association studies of age-related macular degeneration and Crohn’s disease have both revealed new mechanisms of disease that are now being pursued as therapeutic targets.
In the final NEJM Perspective, Peter Kraft and David Hunter from the Harvard School of Public Heath address genomewide association studies from the point of view of risk prediction. They argue that while reliable risk predictions may be possible someday, testing for susceptibility based on common variants is currently premature for most conditions. Like Goldstein, they note the small effect sizes associated with many of the common variants found so far.
“These factors suggest that many, rather than few, variant risk alleles are responsible for the majority of the inherited risk of each common disease….Estimates of risk based on established locus associations are therefore likely to change substantially in the next few years [as more variants are found],” Kraft and Hunter write.
Hirschhorn has a more optimistic view of the value of the variants identified through genomewide association studies. He maintains throughout his article that the point of doing these studies is not risk prediction, but does say that it is likely that for some diseases, useful predictive information will emerge. In fact, for some diseases the variants found to-date already give as much information to clinicians as other routinely used measures.
“For several diseases, associated variants already explain 10 to 20% or more of heritability, a magnitude that is similar to the proportion of risk explained by nongenetic tests in widespread clinical use (such as levels of low-density lipoprotein cholesterol or prostate-specific antigen),” Hirschhorn writes.
Hirschhorn goes on to argue that it is not how of much a disease’s genetics a variant or collection of variants explains, but how this information can shift the cost-benefit ratio of available clinical interventions.
“For diseases without potential therapies, even perfect prediction might not be clinically useful. By contrast, for diseases with effective preventive measures that are too costly or for which the risk-benefit balance is nearly neutral, small increments in predictive power could help effectively target preventive efforts, with substantial clinical impact,” he writes.
Looking back at that 2005 bet between Collins and Gelehrter, it’s clear that not even the biggest names in the field could imagine just how far human genetics would come in such a short amount of time. There’s no disputing that the study of common DNA variants has pushed science forward. And as new technologies and methods are developed, we’re sure to see even more progress. We here at 23andMe hope to be a part of some of those new discoveries.